hypgraph 0.8.2__tar.gz → 0.8.3__tar.gz

{hypgraph-0.8.2 → hypgraph-0.8.3}/CHANGELOG.md

@@ -1,5 +1,14 @@
 # Changelog
 
+## 0.8.3 — 2026-11-03: the next-experiments view in a stable order
+
+- **`verify --standing` no longer fails on a sound bundle with two or more untested hypotheses.** `next_experiments`
+  listed `untested_hypotheses` by iterating a set of ids, so the order followed Python's per-process string hashing:
+  a bundle exported in one process carried one order and its re-derivation another, and `hypgraph verify --standing`
+  reported a standing mismatch (`<inv>/next`) on a record whose content was identical. The list now follows the order
+  the hypotheses were added. Found on a 785-node record; that record and three smaller ones now derive identically
+  under six hash seeds.
+
 ## 0.8.2 — 2026-11-01: an extreme that quotes its own counterexample; a disputed answer
 
 - **A quoted counterexample no longer excuses an extreme.** A stated count used to exempt an extreme whenever its

{hypgraph-0.8.2 → hypgraph-0.8.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: hypgraph
-Version: 0.8.2
+Version: 0.8.3
 Summary: The evidence layer for AI research agents: a hash-chained research record that checks every claim as it is written and derives how well each is known.
 Project-URL: Homepage, https://hypgraph.com
 Project-URL: Documentation, https://hypgraph.com/engineers/

{hypgraph-0.8.2 → hypgraph-0.8.3}/docs/PLAN.md

@@ -109,7 +109,7 @@ claim's derived standing.
 Multi-writer sync between graphs (bundles cover sharing), authentication (the host's job; the HTTP server binds to
 loopback and is read-only unless started with `--writable`), and a hosted service.
 
-## Status at 0.8.2 (2026-11-01)
+## Status at 0.8.3 (2026-11-03)
 
 All seven phases are done: the library, bundles, the server and UI, the MCP server, the website, Pegasus running on
 the package, and releases to PyPI from tags with trusted publishing. What came after the plan, by release, is in

{hypgraph-0.8.2 → hypgraph-0.8.3}/docs/RELEASE-READINESS.md

@@ -256,6 +256,11 @@ was taken from a changelog without being checked against the code.*
 > claim writes changes 5, all now refused (4 lies, 1 retraction worded as the claim). #106 is merged with green CI.
 > Released as 0.8.2, with the `dispute` demo (#107).
 
+> **After 0.8.2 (2026-11-03).** One determinism fix in a derived view, no change to the checks: `next_experiments`
+> listed untested hypotheses in a set's hash order, so `verify --standing` could report a mismatch on a sound bundle
+> with two or more of them when the exporting and verifying processes hashed strings differently. The list now follows
+> the order the hypotheses were added. #109 is merged with green CI. Released as 0.8.3.
+
 ## Recommendation: CONDITIONAL GO
 
 HypGraph is ready for a public developer preview once #59, #60 and #61 are merged with green CI and released as one

{hypgraph-0.8.2 → hypgraph-0.8.3}/src/hypgraph/__init__.py

@@ -16,7 +16,7 @@ Standing is derived on every read, never stored; every write is an event in a SH
 their numbers; reviews pin versions; independence is counted, not assumed. See https://hypgraph.com.
 """
 
-__version__ = "0.8.2"
+__version__ = "0.8.3"
 
 from .bundle import anchored, verify_bundle  # noqa: E402
 from .graph import Author, Graph, GraphError, file_sha256  # noqa: E402

{hypgraph-0.8.2 → hypgraph-0.8.3}/src/hypgraph/plan.py

@@ -48,7 +48,8 @@ def next_experiments(g: "Graph", inv: str, limit: int = 10) -> dict:
     """Planned experiments ranked by how well they discriminate between live hypotheses, whether they can show their
     own power, and cost — duplicates and leased work flagged. Plus live hypotheses that no experiment tests yet."""
     s = g.snapshot()
-    live = set(_live_hypotheses(s, inv))
+    order = _live_hypotheses(s, inv)   # the record's order; the set is for membership only
+    live = set(order)
     dup = {x for grp in duplicates(s, inv) for x in grp[1:]}
     rows = []
     for i, n in s.nodes.items():
@@ -69,7 +70,7 @@ def next_experiments(g: "Graph", inv: str, limit: int = 10) -> dict:
     rows.sort(key=lambda r: (r["duplicate"], r["lease"] is not None, -r["score"]))
     tested = {e["dst"] for x, n in s.nodes.items() if n["type"] == "experiment" and n["inv"] == inv
               for e in s.out.get(x, []) if e["rel"] == "tests"}
-    gaps = [g.brief(s, h) for h in live if h not in tested]
+    gaps = [g.brief(s, h) for h in order if h not in tested]
     return {"ranked": rows[:limit], "untested_hypotheses": gaps,
             "rule": "score = hypotheses it separates (distinct expected outcomes) × 1.5 if it can show its own power "
                     "÷ log2(2 + cost in minutes); duplicates and leased work last"}

hypgraph-0.8.3/tests/test_next_order.py

@@ -0,0 +1,45 @@
+"""The next-experiments view lists untested hypotheses in the record's order, not the hash order of a set.
+
+``next_experiments`` built ``untested_hypotheses`` by iterating a set of live hypothesis ids. String hashing is salted
+per process, so the order changed from run to run, and ``hypgraph verify --standing`` failed on any bundle with two or
+more untested hypotheses exported in another process: the bundle carried one order, the re-derivation another.
+Invariants: the list follows the order the hypotheses were added; re-deriving a bundle's standing under any hash seed
+agrees with what it carries."""
+import json
+import os
+import subprocess
+import sys
+
+from hypgraph import Graph
+
+TITLES = [f"Candidate {c} explains the readings." for c in "ABCDEFGH"]
+
+
+def record():
+    g = Graph()
+    inv = g.author("curator").investigation("Readings", "Which candidate explains the readings?")["id"]
+    a = g.author("a")
+    hs = [a.add("hypothesis", inv, statement=t, predictions=["it fits the held-out readings"])["id"] for t in TITLES]
+    a.add("experiment", inv, aim="Compare candidates A and B on the held-out readings.",
+          protocol="Fit both candidates to the same held-out readings.", success_criteria="A fit for each.",
+          method="least squares",
+          links=[{"rel": "tests", "to": hs[0], "data": {"expected": "A fits"}},
+                 {"rel": "tests", "to": hs[1], "data": {"expected": "B fits"}}])
+    return g, inv, hs
+
+
+def test_untested_follow_the_record_order():
+    g, inv, hs = record()
+    got = [h["id"] for h in g.next_experiments(inv)["untested_hypotheses"]]
+    assert got == hs[2:]
+
+
+def test_standing_verifies_under_any_hash_seed(tmp_path):
+    g, inv, _ = record()
+    f = tmp_path / "b.json"
+    f.write_text(json.dumps(g.export(inv)), encoding="utf-8")
+    for seed in ("0", "1", "2", "3", "4", "5"):
+        env = {**os.environ, "PYTHONHASHSEED": seed}
+        r = subprocess.run([sys.executable, "-m", "hypgraph", "verify", "--standing", str(f)], cwd=tmp_path, env=env,
+                           capture_output=True, text=True, encoding="utf-8")
+        assert r.returncode == 0, (seed, r.stdout, r.stderr)